Request for Open-Sourcing Cross-Sensor Data Generation Code #12
Description
Excellent work! Would it be possible to open-source the code related to Cross-Sensor Data Generation? I would like to experiment with simulating data that matches my own sensor characteristics.
Additionally, I have several questions that I don't fully understand:
-
Regarding the Pattern Generation component:
If I already know the parameters of a LiDAR sensor (such as number of scan lines, max_range, vertical FOV, horizontal FOV, etc.), would it be possible to directly construct the ray pattern without relying on the pattern generation network? -
Regarding the Data Generation component:
The paper mentions: "With the estimated Ray Pattern, simulation LiDAR point cloud can be generated by leveraging the CG (Computer Graphics) techniques. We project the scene onto 6 faces of a cube centered at the LiDAR center to form the cube maps, the environment point cloud is rendered with surface splatting techniques to obtain the smooth and holeless projection while the obstacle CAD models are rendered with regular methods."
When you mention "project the scene onto 6 faces of a cube," does this refer to projecting a simulation scene onto six faces within the simulator to obtain 6 cube depth maps?
- Alternative application scenario:
If I have two LiDARs (A and B), and I construct the ray patterns for both sensors through the pattern generation component, and I have actual data collected by LiDAR A, could I directly project the A data onto 6 cube maps and then use B's ray pattern to select corresponding points from the cube maps to construct data that matches LiDAR B's characteristics? This approach seems somewhat unreasonable to me.
I would greatly appreciate your insights on these questions. Thank you!